duygut / Dealing_with_Imbalanced_Data

This project is about how you can deal with imbalanced data and which performance metrics' particularly important compared to usual practices fairly balanced data.

Kaggle - Credit Card Fraud Detection

This repository is IPython Notebook for Kaggle Dataset, https://www.kaggle.com/mlg-ulb/creditcardfraud

• The aim of this project is detecting fraudulent or non-fraudulent transactions while dealing with imbalanced data. To achieve this, various supervised learning algorithms will be used and the results will be compared.

• Imbalanced data refers to classification problems based on the binary class inequality. There are several methods for dealing with this problem like Re-Sampling, Generate Synthetic Samples, Anomaly Detection Methods or performance metrics instead of accuracy results.

• In this project, the undersampling method will be implemented to the majority class and performance metrics such as Precision, Recall, F1 Score and AUC and some anomaly detection methods like one-class SVM and Neural Network will be used to find the best algorithm which highly predicted fraudulent or non-fraudulent transactions.

Dataset can be found in below link

https://www.kaggle.com/mlg-ulb/creditcardfraud

The project has 4 main topics:

1. Data Exploration
2. Hyperparameter Optimisation
3. Model Building
4. Comparing Performance Metrics

Dependencies

• Pandas
• NumPy
• SciKit-Learn
• Seaborn
• Matplotlib